OS/2 Professional/OS2PRO194.ISO/os2/packer/infozip/sources/vms.zoo/VMS.notes

Text File | 1992-02-29 | 13.1 KB | 307 lines

                     VMS Notes for UnZip 4.2
                        28 February 1992


The various VMS tweaks to UnZip 4.2 and ZipInfo 0.96 were tested on a
VAX 8600 running VMS 5.2 (and, later, VMS 5.4) and VAX C 3.0.  Older
versions were also tested on a VAX 11/785.

To build UnZip (and its trusty sidekick, ZipInfo), just run the included
command file MAKE_UNZIPINFO.COM (i.e., "@make_unzipinfo").  By default,
this creates shareable-image executables, which are smaller and (sup-
posedly) load faster than the normal type.  They also (supposedly) will
be better able to take advantage of any bug fixes or new capabilities
that DEC might introduce, since the library code isn't built into the
executable.  The shared executable is about a quarter the size of the
ordinary type in the case of UnZip.

[Btw, the VMS make utility "MMS" seems not to be compatible enough with
Unix make to use the same makefile.  Antonio Querubin, Jr., sent along an
MMS makefile, subsequently modified by Igor Mandrichenko.  Read the
comments at the top of DESCRIP.MMS for more info.]

UnZip is written to return the standard PK-type return codes (or error
codes, or exit codes, or whatever you want to call them).  Unfortunately,
VMS insists on interpreting the codes in its own lovable way, and this
results in endearing commentary such as "access violation, error mask =
0005, PC = 00003afc" (or something like that) when you run UnZip with no
arguments.  To avoid this I've added a special VMS_return() function which
either ignores the error codes (and exits with normal status) or interprets
them, prints a semi-informative message (enclosed in square [] brackets),
and then exits with a normal error status.  I personally can't stand the
latter behavior, so by default the error codes are simply ignored.  Tastes
vary, however, and some people may actually like semi-informative messages.
If you happen to be one of those people, you may enable the messages by
recompiling misc.c with RETURN_CODES defined.  (This adds a block or two
to the executable size, though.)  The syntax is as follows:
    cc /def=(RETURN_CODES) misc

To use UnZip in the normal way, define a symbol "unzip" as follows:
    unzip :== "$diskname:[directory]unzip.exe"
(substitute for "diskname" and "directory" as appropriate, and DON'T FORGET
THE `$'!  It won't work if you omit that.)  In general it's wise to stick
such assignments in your LOGIN.COM file and THEN forget about them.  It is
no longer necessary to worry about the record type of the zipfile...er,
well, most of the time, anyway (see the Kermit section below).

Having done all this you are ready to roll.  Use the unzip command in
the usual way; as with the Unix, OS/2 and MS-DOS versions, this one uses
'-' as a switch character.  If nothing much happens when you do a directory
listing, for example, you're probably trying to specify a filename which
has uppercase letters in it...VMS thoughtfully converts everything on the
command line to lowercase, so even if you type:
    unzip -v zipfile Makefile
what you get is:
    unzip -v zipfile makefile
which, in my example here, doesn't match the contents of the zipfile.
This is relatively easy to circumvent, by enclosing the filespec(s) in
quotes:
    unzip -tq unzip401 "Makefile" "VMS*" *.c *.h
[This example also demonstrates the use of wildcards, which act like Unix
wildcards, not VMS ones.   In other words, "VMS*" matches files VMSNOTES,
VMS_MAKE.COM, and VMSSHARE.OPT, whereas the normal VMS behavior would be
to match only the first file (since the others have extensions--ordinarily,
you would be required to specify "VMS*.*").]

Note that created files get whatever default permissions you've set, but
created directories additionally inherit the (possibly more restrictive)
permissions of the parent directory.  And, of course, things probably won't
work too well if you don't have permission to write to whatever directory
into which you're trying to extract things.  (That made sense; read it
again if you don't believe me.)

ZipInfo, by the way, is an amusing little utility which tells you all sorts
of amazingly useless information about zipfiles.  Occasionally it's useful
to debug a problem with a corrupted zipfile (for example, we used it to
find a bug in PAK-created zipfiles, versions 2.5 and earlier).  Feel free
to blow it away if you don't need it.  It's about 30 blocks on my system,
and I find I actually prefer its listing format to that of UnZip now (hardly
surprising, since I wrote it :-) ).  I also find it useful to use "ii"
rather than "zipinfo" as the symbol for zipinfo, since it's easier to type
than either "zipinfo" or "unzip -v", and it echoes the common Unix alias
"ll" for the similar style of directory listings.  Oh, and the reason it's
still got a beta version number is that I haven't finished it yet--it would
be better with an automatic paging function, for example.  Oh well.

RANDOM OTHER NOTES:  (1) Igor Mandrichenko (leader of our fearless Russian
contingent) rewrote major portions of the VMS file-handling code, with
the result that UnZip is much smarter now about VMS file formats.  He
made a corresponding set of changes to Zip, but those are not yet avail-
able.  When they are, however, UnZip will be able handle them.  In the
meantime, the notes at the end of this file from Hugh Schmidt and Mike
Freeman give hints on how to save VMS attributes using Zip 1.0.  These
notes refer to UnZip 4.1, but they should still hold for this version.
(2) Zip 1.0 cannot handle any zipfile that isn't in stream-LF format, so
you may need to use Rahul Dhesi's BILF utility which is included with
UnZip.  It may also be necessary for certain other special occasions, like
when you've forgotten to set the correct Kermit parameters while uploading
a zipfile (see Hugh Schmidt's note below for comments about Kermit, too).

Greg Roelofs,
 not really involved with VMS UnZip anymore, but updating this file one
 last time since it's mine. :-)

====================

From INFO-ZIP Digest (Wed, 6 Nov 1991), Volume 91, Issue 290

                VMS attributes and PKZIP compatibility
                  VMS attributes restored! (2 msgs)

------------------------------

Date: Tue, 5 Nov 91 15:31 CDT
From: Hugh Schmidt <[e-mail]>
Subject: VMS attributes and PKZIP compatibility
Message-ID: <[e-mail]>

           ******************************************************
(1)        *** Retaining VMS attributes - a proposed strategy ***
           ******************************************************

This is a proposed strategy for recovering VMS file attributes after
zip/unzip:

a) capture VMS file attributes: VMS ANALYZE/RMS/FDL/OUTPUT=fdlfile vmsfile.ext
b) compress files on VMS......: ZIP zipfile vmsfile.ext, fdlfile.fdl
c) uncompress files on VMS....: UNZIP zipfile vmsfile.ext, fdlfile.fdl
d) recover VMS file attributes: CONVERT/FDL=fdlfile.fdl vmsfile.ext vmsfile.ext

The wrinkle is that UNZIP creates files which are unreadable by CONVERT
despite a concerted effort to accomodate VMS file management system:

file_io.c, line 178: ...creat(filename,0, "rat=cr", "rfm=streamlf")

These files are unCONVERTABLE because they appear to VMS to contain
records with 512+ bytes.  Poring over VMS manuals (Programming V-6A, VAX
C RTL Reference Manual) doesn't readily suggest better alternatives.
Experimentation with "rat=fix", etc. may help suppress the spurious
block-end delimeters.

          ****************************************************
(2)       *** VMS ZIP and PKZIP compatibility using KERMIT ***
          ****************************************************

Many use Procomm's kermit to transfer zipped files between PC and VMS
VAX.  The following VMS kermit settings make VMS-ZIP compatible with
PKZIP:

                                             VMS kermit      Procomm kermit
                                           ---------------   --------------
Uploading PKZIPped file to be UNZIPped:    set fi ty fixed    set fi ty bi
Downloading ZIPped file to be PKUNZIPped:  set fi ty block    set fi ty bi

"Block I/O lets you bypass the VMS RMS record-processing capabilities
entirely", (Guide to VMS file applications, Section 8.5).  The kermit
guys must have known this!

         ************************************************
(3)      *** Making ZIP files compatible with VMS RMS ***
         ************************************************

It looks like a crummy patch, but to make VMS RMS happy, I inserted the
following near line 570 in the putend() section of zipfile.c:

#ifdef VMS
 fprintf(f,"\n")
#endif

It's probably of no consequence, but it satisfies VMS ANALYZE.

------------------------------

Date: Tue, 5 Nov 91 19:40 CDT
From: Hugh Schmidt <[e-mail]>
Subject: VMS attributes restored!
Message-ID: <[e-mail]>

           ************************************************************
           ***   Aha!  It works!  VMS attributes can be restored! *****
           ***                                                      ***
           *** Change FAB$C_STMLF to FAB$C_FIX in file_io.c         ***
           *** ANALYZE => .FDL | CONVERT => original VMS attributes ***
           ************************************************************

(1) Change line 147 in file_io.c and dropping lines 148-149:

    fileblk.fab$b_rfm = FAB$C_STMLF;    /* stream-LF record format */
    fileblk.fab$b_rat = FAB$M_CR;       /* carriage-return carriage ctrl */
    /*                      ^^^^ *NOT* V_CR!!!     */

=>

    fileblk.fab$b_rfm = FAB$C_FIX;    /* fixed record format */

(2) Use VMS ANALYZE to store the VMS attributes for the original file
    into a file descripion language file.  Save the .FDL file with the
    original file.

(3) Apply the original attributes to the UNZIPped file using VMS CREATE
    and the description file (test.fdl).

In the following example, the original attributes of a SPSSX system
format file were restored.  Later, SPSSX successfully read the UNZIPped/
CONVERTed file.

!******** Procedure (or method for you ai'ers)
!********
!******** Create an .FDL file using ANALYZE

$ analyze/rms/fdl/out=test test.spssxsav;1

!******** ZIP the file and its description file

$ zip test test.spssxsav;1 test.fdl;1
adding test.spssxsav (imploded 62%)
adding test.fdl (imploded 90%)

!******** When the ZIPPED file is needed, UNZIP it and its description file

$ unzip test
  Exploding: test.spssxsav
  Exploding: test.fdl

!******** To restore the original attributes, apply the description file to
!********  the UNZIPped file using VMS CONVERT

$ convert/fdl=test  test.spssxsav;2 test.spssxsav;3

! ******* The following show that the VMS attributes are restored.

$ analyze/rms/check test.spssxsav;1

    File Spec: PROJDISK5:[322042.CORE]TEST.SPSSXSAV;1
    File Organization: sequential
    Record Format: fixed
    Record Attributes:
    Maximum Record Size: 4
    Longest Record: 4

The analysis uncovered NO errors.

$ analyze/rms/check test.spssxsav;2
    File Spec: PROJDISK5:[322042.CORE]TEST.SPSSXSAV;2
    File Organization: sequential
    Record Format: undefined
    Record Attributes:
    Maximum Record Size: 0
    Longest Record: 512
The analysis uncovered NO errors.

$ analyze/rms/check test.spssxsav;3
    File Spec: PROJDISK5:[322042.CORE]TEST.SPSSXSAV;3
    File Organization: sequential
    Record Format: fixed
    Record Attributes:
    Maximum Record Size: 4
    Longest Record: 4

Thank you all for your help and comments.  It was very reasurring to
know that I was on the right track.  This is a humble contribution
compared to the great efforts of others in the ZIP project.

------------------------------

Date: Wed, 6 Nov 91 12:49:33 EST
From: Mike Freeman <[e-mail]>
Subject: Re: VMS attributes restored!
Message-ID: <[e-mail]>

Hello, all.

[e-mail] has come up with a commendable method for saving/
restoring record attributes in Vax/VMS files when ZIPping/UNZIPping.
There are a couple of potential problems in his specific implementation,
however:

(a) It would be better to leave the output file(s) in STREAM_LF format
    and use Joe Meadows' FILE utility to put them in fixed-record
    format if CONVERT has trouble reading them.

    Outputting a text file as fixed-record format could be a pain, and
    one would have to use FILE anyway.

    (Incidentally, I ZIPped up a text file, UNZIPped it and used Hugh's
    method to restore the attributes.  CONVERT had no trouble with the
    STREAM_LF format -- it's only a problem with binary files and, as I
    say, Joe Meadows' FILE will fix this.)

(b) Even if ANALYZE/RMS complains, I do not think one ought to put the
    "\n" in as Hugh advocates.  This contradicts the ZIP philosophy of
    byte-for-byte storage/retrieval.

    As Mr. Roelofs has said, there is nothing wrong with the ZIP file;
    it's just that ANALYZE/RMS expects STREAM_LF files to end in an
    <LF>.

Anyway, I was able to use Hugh's method without modifying ZIP 1.0.  We
really ought to get the word out about FILE, though -- it's very handy.

Mike Freeman, K7UIJ     |       Internet: [e-mail]
301 N.E. 107th Street   |       GEnie: M.FREEMAN11
VancouvEr, WA 98685 USA |       Confidence is the feeling you have
Telephone [phone] |       before you understand the situation.

------------------------------